Know the truth before you vote

What we did: we took the live 2026 manifesto pages on 15 October 2026 and recorded the counts as they stood. We read the manifesto data, not the parties’ own words, for this introduction, and we make no new claim about any party. Sources are listed at the end.

Why a campaign, and why now #

A manifesto is a promise made in public. In most countries, a voter reading one cannot easily tell which promises belong to the office on the ballot paper and which belong to someone else. A mayor can promise to fix a water pipe. A mayor cannot promise to change the immigration law. Between those two sit many promises that are harder to place.

South Africa is a good place to test this because its Constitution writes down what a municipality may do. It speaks of three spheres of government: national, provincial and local. Each has its own powers. The list for a municipality is short and public, which gives a voter and a fact-checker something fixed to check against.

Ukweli is a content integrity platform. Its day job is to read a document and say which of its claims, quotes and sources cannot be confirmed. A manifesto is a document full of claims. The campaign applies the same discipline to the promises that are about to be put to a vote.

The landing page carries a live count of the time left until voting day, 4 November 2026.

What the Ukweli LGE 2026 Manifesto campaign is #

LGE stands for local government elections. The campaign is a set of connected pages, plus the related folios.

The landing page gives the headline numbers and the countdown. Each of the twelve parties whose manifestos Ukweli has analysed has its own page, where every promise appears in the party’s own words, with the page number it was taken from, and the reason for Ukweli’s finding. A comparison page lets a reader put parties side by side on ten themes, such as safety or housing, without anyone being told who did better. And the folios, of which this is the first on the campaign, step back and explain what the numbers mean. Between now and the vote on 4 November 2026, we intend to analyse the curated data in more depth and publish further LGE 2026 Manifesto folios on what it shows.

The first pages of the campaign went live on 4 October 2026.

How a promise is read #

Every promise goes through the same steps, and it helps to see them in order.

First, find the promise. Ukweli reads the party’s own manifesto document and picks out each sentence that commits the party to do something. Where one sentence holds several separate promises, Ukweli splits it and reads each part. Some sentences cannot be split without breaking them. Those stay whole.

Second, quote it. The promise is shown in the party’s own words, with its page. The quotation marks show exactly where the party’s words start and stop.

Third, check it against the law. Ukweli compares the promise with the Constitution’s lists of what a municipality may administer, and with the national laws that explain those lists. An AI model does the first reading. A finding is published only when the readings agree.

Fourth, say so when it cannot tell. If the readings do not agree, or the sentence is too short or vague to test, the card is marked "Could not be assessed by Ukweli". Ukweli publishes no verdict rather than a wrong one. That label is not a finding about the promise or about the party. It is a statement about the reading.

The outcome of that reading is one of six groups. These are the live counts on 15 October 2026.

One caution comes with those counts. They count promises as written. One long promise and one short promise each count once, and the count says nothing about how important a promise is. They also add together twelve manifestos on purpose, because the campaign does not compare one party’s result with another’s.

Ukweli reads a document before it goes out and tells you which of its claims, quotes and sources cannot be confirmed. Put a document through it →

Who is in, and who is not yet #

Twelve manifestos are analysed, shown here in alphabetical order with the number of promises Ukweli found in each.

Two points explain the list.

Selection is about documents, never about parties. Ukweli analyses any 2026 local government manifesto it can obtain from the party’s own publication. No party is in or out because of its size, its coverage, or anyone’s view of how it is run. An absence is a recorded state with a reason.

"Reported" means reported. Ukweli counts a manifesto launch only when a dated report says the launch happened. A promise to launch, an interview, or a manifesto from an earlier election is not a launch. A nightly check looks for new launches, and the number on the landing page moves when one is confirmed. Five parties have reported a launch, but Ukweli cannot yet access a readable manifesto document online for them.

Other parties are contesting the elections. This campaign covers only the manifestos it holds.

Each party’s assessment is one click away.

ACDPRead the assessment →ActionSARead the assessment →AMMRead the assessment →ANCRead the assessment →BOSARead the assessment →DARead the assessment →EFFRead the assessment →FF PlusRead the assessment →GOODRead the assessment →IFPRead the assessment →SACPRead the assessment →UDMRead the assessment →

Each button opens that party’s assessment on ukweli.io in a new tab, so this page stays open. Party logos belong to the parties and are shown for identification only. Ukweli is not affiliated with any party.

How manifestos are found, and how the field is watched #

Finding a manifesto. Ukweli looks for each party’s own published 2026 manifesto. That document is the only source of the party’s promises. Ukweli does not work from news reports or summaries of a manifesto. When it finds a document, it records where it came from and stores a fingerprint of the file, so any later change can be seen.

Recording an absence. If a party has launched a manifesto but no readable document can be reached, Ukweli records that state and the reason. It does not guess what the manifesto says.

Watching the field. The election changes every week, so we also review what other organisations publish about it. The next section comes from that review, and it carries the date it was done. We do not list where we look or how often. Keeping that private stops anyone from shaping what we see.

Other work on the same election, as at 15 October 2026 #

Ukweli is not alone. Eight other groups publish related checks or tools that we found as at 15 October 2026.

This section describes what each organisation says on its own pages, which we read on 15 October 2026. It is a snapshot, and it will go out of date. The list is not exhaustive: we make no claim that these are the only groups checking or monitoring this election, and others may be working on it that we have not found.

Africa Check is a fact-checking organisation. Its review of 23 September started with five of the eight largest parties in the National Assembly, across seven municipal issues. It now covers six, and says it will add the MK Party and the Patriotic Alliance when their manifestos are published. It notes that some promises are specific enough to measure, while others are broad, lack deadlines or depend partly on national or provincial government. It also tells voters to ask three questions: “Is it actually in the municipality’s power to deliver?”, can the promise be measured, and does the manifesto say how it will be paid for. We agree with all of this. Its 28 September report then tests seven claims in the DA’s manifesto about conditions today and the DA’s record, such as electricity tariffs, refuse collection and the cost of crime. It rated five correct, one understated and one exaggerated. Ukweli does not test whether a claim about today is true. It takes each promise about tomorrow, one at a time, names the item in the Constitution’s municipal lists it is checked against, and shows the working.

Daily Maverick, a news site, runs a series called Manifesto test, by Ferial Haffajee. Each piece takes one party’s manifesto and puts it through ten tests under plain headings, such as "Turn on the taps". It adds a question on coalition strategy and notes where a manifesto gives no target, deadline or budget. The pieces share one template so readers can compare parties. We found pieces on at least six parties, the EFF, DA, SACP, ANC, UDM and IFP, from 10 August to 5 October. Ukweli asks a different question: is the promise within a council’s powers?

My Vote Counts, a civil society group, released its Municipal People’s Analysis of Manifestos (M-PAM) on 5 October. We read the report, which runs to 26 pages. It sets out what communities need from local government, drawing on published data such as Statistics South Africa’s, and the standards in laws such as the Municipal Systems Act. It then restates the promises of four parties, the ANC, DA, EFF and ActionSA, in plain language under six themes, and leaves the reader to judge whether the promises meet the needs. For ActionSA it uses the party’s 2024 national manifesto, as the report says. It says it does not tell anyone how to vote. It builds on the group’s 2024 edition. The report sorts promises by what falls under a municipality. Ukweli goes promise by promise, quoting each in the party’s words and naming the item in the Constitution’s municipal lists it is checked against.

IOL, a news site, announced an elections portal on 13 October. Its Manifesto Comparison Tool lets voters compare parties by topic, 20 topics in all, from water to immigration. For each topic it lists a party’s commitments in bullet points, with the manifesto page or launch speech they came from. We read all 20 topics through the tool’s own coverage grid. It covers seven parties, and marks each topic for each party as detailed, a brief mention or not addressed. Ukweli’s pages quote every promise they found in the party’s own words, and test each against the municipal lists in the Constitution.

VotaMzansi is a voter information site. Its home page does not name who runs it. It offers sourced party profiles and lets a reader put up to five parties side by side on the issues a municipality controls. It says every fact shows its source and that it is not tied to any party. We read its home page only.

Manifesto Deep Dive (electai.co.za) is a question-and-answer site. Its footer credits a single developer. A reader asks what the manifestos say about an issue, and the answers quote the manifestos, with what the site calls AI-guided context in between. It lists summaries for 76 parties, and each point links to its page in the manifesto. It marks five of them as the party’s 2024 national manifesto, not a 2026 local one. Parties and independent candidates can upload a manifesto for review. It also offers the candidates in each ward, two to four parties side by side, and a quiz that matches a reader’s priorities. We read its home page, its list of summaries and one party’s summary. Ukweli works only from each party’s own published document, so it has not used these copies.

Moxii Africa, a media monitoring group, runs Election Watch with SANEF (the South African National Editors’ Forum) and Africa Check. Two dashboards on its Election Watch page are live: Dexter, which shows how media cover parties and issues, and PAdRe, a repository of political adverts and the money spent on them. It lists a Promise Tracker as coming soon, where candidates share commitments so voters can hold them to their word after the vote.

People’s Assembly, a site that tracks the work of elected representatives, posted a guide to manifestos, independent candidates and community manifestos on 16 September.

The table below places each by what it reads and what the reader gets. It has no ticks or crosses, because it is not a scorecard.

The difference is the test. Ukweli quotes every promise it found in the party’s own words, with the page, and checks each one. It names the item in the Constitution’s municipal lists that the promise is checked against, so a reader can check it without trusting Ukweli. It gives no score and no rank. At the date of this folio it covers twelve manifestos.

What the campaign will not do #

Three rules were set at the start and they apply to every page and every folio.

It will not score, grade, rank or endorse. The landing page says it plainly: Ukweli does not score, rank or endorse any party. There is no league table, no winner and no loser.

It will not say whether a promise will be kept. The page says that Ukweli does not say whether a promise will be kept, whether it is affordable, or whether a party’s figures are right. Those are questions for the voter and the candidate.

It will not treat "outside" as "false". A promise outside a council’s powers is not a lie. A council can still lobby, partner with or press the government that holds the power. A voter who knows whose power it is can ask the candidate how, and not only whether.

There is a fourth limit, and it is about Ukweli itself. An AI model reads the promises, and the readings can be wrong. That is why a finding is published only when the readings agree, and why a case where they do not agree is labelled "Could not be assessed by Ukweli". If you find an error, write to hello@ukweli.io.

How to use it #

There are three ways in, depending on what you want.

If you are a voter, start with your party. Open its page, read the promises that matter to you in its own words, and look at the reason Ukweli gives. Then take the question to the next candidate who knocks on your door: whose power is this, by when, and with whose money?

If you are comparing, use the comparison page. Choose the parties you want to see and read their promises one theme at a time. The page shows the words. It does not show a winner.

If you are a journalist, a researcher or a candidate, read the method. Folio 08, "Whose power is it?", sets out how the readings were made, which laws were read, and what the first findings were. It is the deep companion to this introduction.

What we have not yet measured #

A method is only as honest as its stated limits. Five stand out.

Missed promises. We have not measured how many promises the reading misses. The counts are counts of promises found, not of everything a party promised.

Split promises. Where one sentence holds several promises, Ukweli splits it, and each piece gets its own result. This lifts the count above the number of sentences.

Hard-to-read pages. Some manifestos include pages that are pictures of text with no real text behind them. Those pages are harder to read.

Changing counts. The counts move when a manifesto is added. They moved when the BOSA and GOOD manifestos were added, and when the reading was re-run on a newer version.

Who checks Ukweli. Ukweli publishes its working so that readers can check it. If you find an error, write to hello@ukweli.io. We correct it and say so.

What this introduction leaves out #

This folio introduces the campaign. It does not report new findings about any party. The numbers above are counts of promises as the pages show them on 15 October 2026. They move when a manifesto is added or a finding is corrected, and this folio does not change with them. The section on other work is a snapshot as at 15 October 2026 and will move too.